

- Aug 17
- 14 min
Unveiling "Continuous Diffusion for Categorical Data Estimation for LLMs"
The CDCD framework harmoniously integrates continuous and discrete data generation methodologies, capitalizing on their respective strengths


- Jun 9
- 13 min
Exploring Geometric Deep Learning and its Impactful Applications
The key idea behind geometric deep learning is to develop neural network architectures that can operate directly on geometric data.


- May 11
- 14 min
Understanding CLIP
The reason why CLIP is so effective in zero-shot learning is because of the contrastive learning approach to jointly train a neural network.